def setUp(self):
""" The setup consists of four particles placed at the
vertices of a unit square. The force function to be tested is:
..math::
f_i = \sum_{j=1}^{4} \frac{m_j}{|x_j - x_i|^3 +
\eps}(x_j - x_i)
The mass of each particle is 1
"""
self.precision = "single"
self.np = 4
x = numpy.array([0, 0, 1, 1], numpy.float64)
y = numpy.array([0, 1, 1, 0], numpy.float64)
z = numpy.zeros_like(x)
m = numpy.ones_like(x)
tmpx = numpy.zeros_like(x)
tmpy = numpy.zeros_like(x)
tmpz = numpy.zeros_like(x)
self.pa = pa = base.get_particle_array(name="test",
x=x,
y=y,
z=z,
m=m,
tmpx=tmpx,
tmpy=tmpy,
tmpz=tmpz,
cl_precision=self.precision)
self.func = func = sph.NBodyForce.get_func(pa, pa)
self.eps = func.eps
if solver.HAS_CL:
self.ctx = ctx = cl.create_some_context()
self.q = q = cl.CommandQueue(ctx)
pa.setup_cl(ctx, q)
pysph_root = solver.get_pysph_root()
template = solver.cl_read(
path.join(pysph_root, "sph/funcs/external_force.clt"),
function_name=func.cl_kernel_function_name,
precision=self.precision)
prog_src = solver.create_program(template, func)
self.prog = cl.Program(ctx,
prog_src).build(solver.get_cl_include())
def setUp(self):
""" The setup consists of four particles placed at the
vertices of a unit square. The force function to be tested is:
..math::
f_i = \sum_{j=1}^{4} \frac{m_j}{|x_j - x_i|^3 +
\eps}(x_j - x_i)
The mass of each particle is 1
"""
self.precision = "single"
self.np = 4
x = numpy.array([0, 0, 1, 1], numpy.float64)
y = numpy.array([0, 1, 1, 0], numpy.float64)
z = numpy.zeros_like(x)
m = numpy.ones_like(x)
tmpx = numpy.zeros_like(x)
tmpy = numpy.zeros_like(x)
tmpz = numpy.zeros_like(x)
self.pa = pa = base.get_particle_array(name="test", x=x, y=y, z=z,
m=m, tmpx=tmpx, tmpy=tmpy,
tmpz=tmpz,
cl_precision=self.precision)
self.func = func = sph.NBodyForce.get_func(pa, pa)
self.eps = func.eps
if solver.HAS_CL:
self.ctx = ctx = cl.create_some_context()
self.q = q = cl.CommandQueue(ctx)
pa.setup_cl(ctx, q)
pysph_root = solver.get_pysph_root()
template = solver.cl_read(
path.join(pysph_root, "sph/funcs/external_force.clt"),
function_name=func.cl_kernel_function_name,
precision=self.precision)
prog_src = solver.create_program(template, func)
self.prog = cl.Program(ctx, prog_src).build(solver.get_cl_include())
def setup_cl(self):
pa = self.pa
if solver.HAS_CL:
self.ctx = ctx = cl.create_some_context()
self.q = q = cl.CommandQueue(ctx)
pa.setup_cl(ctx, q)
pysph_root = solver.get_pysph_root()
template = solver.cl_read(
path.join(pysph_root, "sph/funcs/pressure_funcs.clt"),
function_name=self.grad_func.cl_kernel_function_name,
precision=self.precision)
prog_src = solver.create_program(template, self.grad_func)
self.prog=cl.Program(ctx, prog_src).build(solver.get_cl_include())
ctx = cl.Context(devices)
q = cl.CommandQueue(ctx, device)
mf = cl.mem_flags
xbuf = cl.Buffer(ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=x)
ybuf = cl.Buffer(ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=y)
zbuf = cl.Buffer(ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=z)
args = (ybuf, zbuf)
pysph_root = solver.get_pysph_root()
src = solver.cl_read(path.join(pysph_root, 'solver/cl_common.cl'),
precision='single')
prog = cl.Program(ctx, src).build(options=solver.get_cl_include())
# launch the OpenCL kernel
prog.set_tmp_to_zero(q, (16, 16, 16), (1,1,1), xbuf, *args)
# read the buffer contents back to the arrays
solver.enqueue_copy(q, src=xbuf, dst=x)
solver.enqueue_copy(q, src=ybuf, dst=y)
solver.enqueue_copy(q, src=zbuf, dst=z)
for i in range(np):
assert x[i] == 0.0
assert y[i] == 0.0
assert z[i] == 0.0
print "OK"
ctx = cl.Context(devices)
q = cl.CommandQueue(ctx, device)
mf = cl.mem_flags
xbuf = cl.Buffer(ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=x)
ybuf = cl.Buffer(ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=y)
zbuf = cl.Buffer(ctx, mf.READ_WRITE | mf.COPY_HOST_PTR, hostbuf=z)
args = (ybuf, zbuf)
pysph_root = solver.get_pysph_root()
src = solver.cl_read(path.join(pysph_root, 'solver/cl_common.cl'),
precision='single')
prog = cl.Program(ctx, src).build(options=solver.get_cl_include())
# launch the OpenCL kernel
prog.set_tmp_to_zero(q, (16, 16, 16), (1, 1, 1), xbuf, *args)
# read the buffer contents back to the arrays
cl.enqueue_copy(q, src=xbuf, dest=x).wait()
cl.enqueue_copy(q, src=ybuf, dest=y).wait()
cl.enqueue_copy(q, src=zbuf, dest=z).wait()
for i in range(np):
assert x[i] == 0.0
assert y[i] == 0.0
assert z[i] == 0.0
print "OK"